Duke-UNC Brain Imaging and Analysis Center
BIAC Forums | Profile | Register | Active Topics | Members | Search | FAQ
Username:
Password:
Save Password   Forgot your Password?
 All Forums
 Support Forums
 Analysis Software Support
 FSL - cope1 report log errors
 New Topic  Reply to Topic
 Printer Friendly
Author Previous Topic Topic Next Topic  

Bethany
BIAC Alum

37 Posts

Posted - May 30 2008 :  3:08:33 PM  Show Profile  Reply with Quote
I have a question about my FSL third level analysis for RiskTime.01. The highest-level report log (the one in the top-level output directory looks fine, and the program did produce data for the contrast (though data that didn't look like what I would have expected based on previous analyses).

However, when I go down one level into the cope1.feat directory and look a the report log there, I get something that looks like this:

Progress Report / Log
Started at Fri May 30 13:36:09 EDT 2008

Higher-level stats

cat ../design.lcon | awk '{ print  }' > design.lcon
/usr/local/fsl/bin/flame --cope=filtered_func_data --vc=var_filtered_func_data --dvc=tdof_filtered_func_data --mask=mask --ld=stats --dm=design.mat --cs=design.grp --tc=design.con  --ols --nj=10000 --bi=500 --se=1 --fm --zlt=100000 --zut=100000
DOF cannot be zero or negative!
DOF cannot be zero or negative!
DOF cannot be zero or negative!
DOF cannot be zero or negative!
DOF cannot be zero or negative!
DOF cannot be zero or negative!
DOF cannot be zero or negative!


The same message (DOF cannot be zero or negative!) repeats for many pages. Then it switches to:

DOF cannot be zero or negative!
DOF cannot be zero or negative!
DOF cannot be zero or negative!
Log directory is: stats
Setting up:
ntptsing=23.000000 

No f contrasts
nevs=1
ntpts=23
ngs=1
nvoxels=221619
Running:
nmaskvoxels=601
njumps = 10000
burnin = 500
sampleevery = 1
nsamples = 9500

Metropolis Hasting Sampling
Number of voxels=601
Percentage done:
 1
stdtr domain error

ndtri domain error

stdtr domain error

ndtri domain error

stdtr domain error

ndtri domain error

stdtr domain error

ndtri domain error

stdtr domain error

ndtri domain error

stdtr domain error

ndtri domain error


Then it keeps counting up the percentage done from 1 to 100 with these same error messages repeated every time, after which it finishes off with the post-stats looking normal.

I don't know what's happening. Is this supposed to be happening? Like I said, it does output a contrast, but none of my older third-levels have log files that contain these sorts of error messages and I can't shake the feeling that so many pages of errors must mean something is wrong.

The second-levels logs all normal (both the top directory report and the report for the individual cope I checked).

Any ideas what's going on?

josh.bizzell
BIAC Staff

USA
118 Posts

Posted - May 30 2008 :  3:19:10 PM  Show Profile  Reply with Quote
I've seen this error before in 3rd level analyses, and it was because of NaN values in one of the copes that was input into the 3rd level.

I detected this by loading the "filtered_func_data" from that 3rd level cope's folder into showsrs2 and found black dots in subjects that were NaN (not a number), usually around the edge of the brain.

If this is the case, you can use fslmaths to convert NaN values to zero with the "-nan" flag.

-Josh
Go to Top of Page

Bethany
BIAC Alum

37 Posts

Posted - May 30 2008 :  3:48:31 PM  Show Profile  Reply with Quote
Yep, that's exactly what mine look like. Thank you!

Does the fact that I have all these NaNs suggest I'm doing something wrong at a lower level of analysis, or is it just FSL moving in mysterious ways?

Am I right in thinking that it's the second-level cope.nii.gz files I need to use fslmaths on?

Thanks again!

Go to Top of Page

josh.bizzell
BIAC Staff

USA
118 Posts

Posted - May 30 2008 :  3:50:46 PM  Show Profile  Reply with Quote
I'm not sure what causes the NaN's in the 2nd level copes.... I was under the assumption that FSL doesn't create NaN values. Maybe someone else knows more.

Yes, you should run fslmaths on the offending subjects' copes with NaN values. Unless you can figure out why the NaNs are being created in the first place (which would be optimal).

- Josh
Go to Top of Page

dvsmith
Advanced Member

USA
218 Posts

Posted - May 30 2008 :  4:24:16 PM  Show Profile  Visit dvsmith's Homepage  Reply with Quote

Are you using empty EVs at any point? I've seen weird things happen when using empty EVs. Sometimes if you just re-run the offending analysis, it will correct itself. (strange... I know)

Go to Top of Page

Bethany
BIAC Alum

37 Posts

Posted - May 30 2008 :  11:39:56 PM  Show Profile  Reply with Quote
I shouldn't be using empty EVs at any point, no. Unless I totally screwed something up somewhere and haven't noticed, which is certainly not impossible.

I got the third levels to run without errors using fslmaths, but got the same (weird) results. Phooey. I'm going to try essentially re-running my original analysis from last year. If that doesn't come out essentially the same, I'll know there's something screwy going on.

Thanks again!

Go to Top of Page
  Previous Topic Topic Next Topic  
 New Topic  Reply to Topic
 Printer Friendly
Jump To:
BIAC Forums © 2000-2010 Brain Imaging and Analysis Center Go To Top Of Page
This page was generated in 0.45 seconds. Snitz Forums 2000